This is an overview of the dataset tidied and uploaded by the NHS R community, which is sourced from NHS England A&E Attendances and Emergency Admissions data. Refer to the previous links for a more detailed description

Get data and tweak

Let’s load the lirbaries first:

We can either download the package and load the data by calling remotes::install_github("https://github.com/nhs-r-community/NHSRdatasets") then data(ae_attendances) or simply:

Let’s take a look at the columns:

It can be useful to check missing values in attendances, breaches or admissions from 2016-04-01 to 2019-03-01:

There’s a big chunk of rows with missing data. Maybe, not all departments or organisaitons started reporting at the same time. But there seems to be more completness from the start of the 2018 financial year.


There are 274 distinct organisation codes in this dataset. This is the Organisation Data Service code, that is an identifier for health providers and other non-NHS organisations issued by NHS Digital.


We can add a bit of extra information to the data set by searching the ODS code in the Organisation Endpoint, that allows to query for a specified code and obtain organisational data. A JSON file is returned with the full organisational data. This endpoint is part of the ODS API Suite.

Here, I use the GET method to query the organisation data by pasting the code with the url, retrieve the content (in JSON format), convert to a list object from JSON, transform to a data frame in long format and get only the fields of post code, town and organisation name. Next, we append all data frames and left join with the data set.

Now let’s add another extra bit: the index of deprivation of the LSOA the organisation belongs to.
First, we link the post code to the LSOA code by using the National Statistics Postcode Lookup UK dataset API by doing something similar to before:

And second, we get the index multiple deprivation lookup and join:

The Indeces of Deprivation ranks every LSOA from in England from 1 (most deprived) to 32844 (least deprived). And because, there is not a definite cutoff that discerns between deprived and no deprived, we can use deciles to get 10 equal groups as recommended in the English Indices of Deprivation 2015 FAQs:

Also, because we will aggregate at local auhtority level, we need to get the boundaries (GeoJSON format in this case) that will be used later:



Plot data

Cummulative attendances

Let’s take a look at the accumulated number of attendances:

top_5_att <- 
  merged_imd %>% 
  select(name, attendances) %>% 
  group_by(name) %>% 
  summarise(sum_att = sum(attendances, na.rm = TRUE)) %>% 
  top_n(5) %>% 
  pull(name)

merged_imd %>% 
  select(org_code, attendances, period) %>% 
  group_by(org_code, period) %>% 
  summarise(att = sum(attendances, na.rm = TRUE)) %>% 
  arrange(period) %>% 
  mutate(csum_att = cumsum(att)) %>% 
  left_join(., select(merged_imd, org_code, name)) %>% 
  distinct() %>% 
  mutate(
    color = case_when(
      name == top_5_att[1] ~ "#b58900",
      name == top_5_att[2] ~ "#cb4b16",
      name == top_5_att[3] ~ "#dc322f",
      name == top_5_att[4] ~ "#6c71c4",
      name == top_5_att[5] ~ "#2aa198",
      TRUE ~ "grey35"
    ),
    alpha = if_else(name %in% top_5_att, 1, 0.2)) %>% 
  ggplot(aes(x = period, y = csum_att, group = org_code)) +
  geom_step(aes(color = color, alpha = alpha)) +
  scale_color_identity(guide = FALSE) +
  geom_text(data = . %>% filter(name %in% top_5_att, period == last(period)) %>% 
              ungroup() %>% 
              arrange(-csum_att) %>% 
              mutate(position = c(1020000, 990000, 960000, 930000, 900000)),
            x = as.Date("2016-04-01"), 
            aes(y = position, label = glue::glue("{str_to_title(name)}: {scales::comma(csum_att)}"), color = color),
            family = "Ubuntu Mono",
            hjust = 0) +
  scale_alpha_identity() +
  scale_x_date(date_labels = "%y-%b", 
               breaks = seq.Date(as.Date("2016-04-01"), as.Date("2019-03-01"), "month"), 
               limits = c(as.Date("2016-04-01"), as.Date("2019-03-01")),
               expand = c(0, 0)) +
  scale_y_continuous(labels = function(x) str_replace(number(x), "\\s0{3}$", "K"), 
                     expand = c(0, 0),
                     limits = c(0, NA)) +
  labs(x = NULL, y = NULL) +
  theme_clean()+
  theme(text = element_text(family = "Ubuntu Mono", color = "gray20"),
        axis.text.x = element_text(angle = 90, vjust = .5),
        plot.background = element_rect(color = "transparent"))

Attendances and breaches of type 1 A&E

Another way to visualize proportions is using geom_ribbon(). We set the x axis as the number of admission, and y axis as number of breaches. If we drew a line perpendicular to the X axis which top limit is any given value of admissions in the X and the bottom is 0, then we could know what % of the X value is any value placed in the vertical line:

ribbon_df <- 
  merged_imd %>%
  select(type, org_code, attendances, breaches, period) %>% 
  filter(type == "1") %>% 
  group_by(org_code, type) %>% 
  summarise(attendances = sum(attendances),
            breaches = sum(breaches))

ribbon1 <- tibble(x = seq(0, max(ribbon_df$attendances)),
                  ymax = x,
                  ymin = seq(0, max(ribbon_df$attendances)) * .75)

ribbon2 <- tibble(x = seq(0, max(ribbon_df$attendances)),
                  ymax = x,
                  ymin = seq(0, max(ribbon_df$attendances)) * .5)

ribbon3 <- tibble(x = seq(0, max(ribbon_df$attendances)),
                  ymax = x,
                  ymin = seq(0, max(ribbon_df$attendances)) * .25)

ribbon4 <- tibble(x = seq(0, max(ribbon_df$attendances)),
                  ymax = x,
                  ymin = 0)

ribbon_df %>%
  ggplot(aes(y = breaches, x = attendances)) +
  geom_ribbon(inherit.aes = FALSE, data = ribbon4, aes(x = x, ymin = ymin, ymax = ymax), alpha = .3, fill = "#c0d8c1") +
  geom_ribbon(inherit.aes = FALSE, data = ribbon3, aes(x = x, ymin = ymin, ymax = ymax), alpha = .3, fill = "#d8d8ab") +
  geom_ribbon(inherit.aes = FALSE, data = ribbon2, aes(x = x, ymin = ymin, ymax = ymax), alpha = .3, fill = "#efcf97") +
  geom_ribbon(inherit.aes = FALSE, data = ribbon1, aes(x = x, ymin = ymin, ymax = ymax), alpha = .3, fill = "#efb7b3") +
  geom_point(alpha = .8, shape = 21, color = "#5D2466") +
  annotate("text", x = 680000, y = 600000, fontface = "bold", hjust = 0, label = "75% - 100%", alpha = .6) +
  annotate("text", x = 700000, y = 450000, fontface = "bold", hjust = 0, label = "50% - 75%", alpha = .6) +
  annotate("text", x = 700000, y = 280000, fontface = "bold", hjust = 0, label = "25% - 50%", alpha = .6) +
  annotate("text", x = 700000, y = 50000, fontface = "bold", hjust = 0, label = "0% - 25%", alpha = .6) +
  expand_limits(x = c(0, 0)) +
  coord_cartesian(ylim=c(0, max(ribbon_df$attendances) * .75)) +
  scale_y_continuous(position = "right", labels = function(x) str_replace(number(x), "0{3}$", "K"), expand = expand_scale(0)) +
  scale_x_continuous(labels = function(x) str_replace(number(x), "0{3}$", "K"), expand = expand_scale(0)) +
  
  labs(x       = "Attendances", 
       y       = "Breaches"
  ) +
  theme_minimal() +
  theme(text = element_text(family = "Ubuntu Mono", color = "gray20"),
        legend.direction = "horizontal",
        legend.key.height = unit(3, "mm"),
        legend.spacing.x = unit(0, "mm"),
        legend.position = c(x = .2, y = .7),
        panel.grid.major = element_blank())

Sesssion info

## R version 3.6.1 (2019-07-05)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 18.04.3 LTS
## 
## Matrix products: default
## BLAS:   /usr/lib/x86_64-linux-gnu/openblas/libblas.so.3
## LAPACK: /usr/lib/x86_64-linux-gnu/libopenblasp-r0.2.20.so
## 
## locale:
##  [1] LC_CTYPE=en_GB.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_GB.UTF-8        LC_COLLATE=en_GB.UTF-8    
##  [5] LC_MONETARY=en_GB.UTF-8    LC_MESSAGES=en_GB.UTF-8   
##  [7] LC_PAPER=en_GB.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_GB.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
##  [1] scales_1.0.0      knitr_1.23        ggthemes_4.2.0   
##  [4] rcartocolor_2.0.0 sf_0.7-7          jsonlite_1.6     
##  [7] httr_1.4.0        forcats_0.4.0     stringr_1.4.0    
## [10] dplyr_0.8.3       purrr_0.3.2       readr_1.3.1      
## [13] tidyr_0.8.99.9000 tibble_2.1.3      ggplot2_3.2.0    
## [16] tidyverse_1.2.1  
## 
## loaded via a namespace (and not attached):
##  [1] tidyselect_0.2.5   xfun_0.8           reshape2_1.4.3    
##  [4] haven_2.1.0        lattice_0.20-38    colorspace_1.4-1  
##  [7] vctrs_0.2.0        generics_0.0.2     htmltools_0.3.6   
## [10] yaml_2.2.0         rlang_0.4.0        e1071_1.7-2       
## [13] pillar_1.4.2       glue_1.3.1         withr_2.1.2       
## [16] DBI_1.0.0          modelr_0.1.4       readxl_1.3.1      
## [19] plyr_1.8.4         lifecycle_0.1.0    munsell_0.5.0     
## [22] gtable_0.3.0       cellranger_1.1.0   rvest_0.3.4       
## [25] codetools_0.2-16   evaluate_0.14      labeling_0.3      
## [28] curl_4.0           class_7.3-15       broom_0.5.2       
## [31] Rcpp_1.0.2         KernSmooth_2.23-15 backports_1.1.4   
## [34] classInt_0.4-1     hms_0.4.2          digest_0.6.20     
## [37] stringi_1.4.3      grid_3.6.1         cli_1.1.0         
## [40] tools_3.6.1        magrittr_1.5       lazyeval_0.2.2    
## [43] crayon_1.3.4       pkgconfig_2.0.2    zeallot_0.1.0     
## [46] xml2_1.2.0         lubridate_1.7.4    assertthat_0.2.1  
## [49] rmarkdown_1.13     rstudioapi_0.10    R6_2.4.0          
## [52] units_0.6-3        nlme_3.1-140       compiler_3.6.1